Possible Duplicate:
Not-significant F but a significant coefficient in multiple linear regression
How can you have a non-significant multiple regression model w/ significant predictors?

My hypothesis was that social engagement would be associated with cognition. I set up a one-way MANOVA in SPSS to look at the association between social engagement (IV, 9 levels) and performance on 6 different cognitive tests. Because the IV had so many levels, I also set up a series of planned contrasts (equivalent to Helmert contrasts) so that I could see where the real difference lay. The overall ANOVAs came back as non-significant for all DVs bar one. However, for 3 of the tests, the same planned contrast came up as significant (p<0.01).

How could the planned contrasts come up as significant if the overall ANOVA was not? And how do I report this? Please help!


marked as duplicate by Macro, Andy W, gung, whuber Aug 25 '12 at 22:00

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.


Planned comparisons are usually powerful but you´ll need to check if Helmert contrast are a type of ORTHOGONAL contrast (independent). If Helmert contrast is in fact a non-orthogonal contrast then you have to adjust the alpha level using Bonferroni, or other procedures, since you are inflating type I error in the comparisons.


Not the answer you're looking for? Browse other questions tagged or ask your own question.